Rose-20B-exl2 / README.md
kingbri's picture
Add README
40e2e6c
metadata
language:
  - en

Information

This is a Exl2 quantized version of Rose-20B

Please refer to the original creator for more information.

Calibration dataset: wikitext

Branches:

  • main: Measurement files
  • 4bpw: 4 bits per weight
  • 5bpw: 5 bits per weight
  • 6bpw: 6 bits per weight

Notes

  • 6bpw is recommended for the best quality to vram usage ratio (assuming you have enough vram).
  • Please ask for more bpws in the community tab if necessary.

Run in TabbyAPI

TabbyAPI is a pure exllamav2 FastAPI server developed by us. You can find TabbyAPI's source code here: https://github.com/theroyallab/TabbyAPI

If you don't have huggingface-cli, please run pip install huggingface_hub.

To run this model, follow these steps:

  1. Make a directory inside your models folder called Rose-20B-exl2

  2. Open a terminal inside your models folder

  3. Run huggingface-cli download royallab/Rose-20B-exl2 --revision 4bpw --local-dir Rose-20B-exl2 --local-dir-use-symlinks False

    1. The --revision flag corresponds to the branch name on the model repo. Please select the appropriate bpw branch for your system.
  4. Inside TabbyAPI's config.yml, set model_name to Rose-20B-exl2 or you can use the /model/load endpoint after launching.

  5. Launch TabbyAPI inside your python env by running python main.py

Donate?

All my infrastructure and cloud expenses are paid out of pocket. If you'd like to donate, you can do so here: https://ko-fi.com/kingbri

You should not feel obligated to donate, but if you do, I'd appreciate it.